Controlling for Average Year of Lots

Another control variable that demonstrated useful for the research is the year of urbanization of the Morphological Neighborhood Area. Québec cadastral system records the date of creation of each cadastral lot. The system was established in late 19th _{century (and rolled out on a city by}

city and neighborhood by neighborhood basis). All previously existing lot were then attributed a creation date corresponding to the creation of the cadastre in their sector, all subsequently created lot were attributed their actual date of creation. Though the data have limitations, this dating system is acceptable for the purposes of this research. The average year of lot creation within the residential patch is negatively correlated with both residential density and street

120

network connectivity variables (Figure 35). These negative correlations demonstrate that earlier residential tissues tend to have higher residential density and better street network connectivity than more recent ones, which is evidence that early residential tissues are more walkable, while more recent residential tissues are predicated on automobility. Furthermore, average year is moderately negatively correlated (r = -0.400) with coverage level and moderately negatively correlated (r = -0.465) with service level. These findings imply that residents living in residential areas urbanized earlier have better access to fresh food at a walking distance than people live in in more recent residential areas. Again, in order to strengthen the interpretation that morphological conditions might indeed explain levels of coverage and service, average year of patches was then controlled in the following correlation analyses.

Figure 32 and 33 showed that inner city areas offer better conditions than suburban areas in terms of residential density, street network connectivity and service level. Figure 39 presents a map, which classifies of residential patches based on average year of creation of their lots. It is useful to remember that before the 1870’s, the city was compact and essentially travelled by foot for daily errands. Horse-drawn streetcars were introduced in the 1860’s that were replaced by an electrified version starting in the early 1890’s. Streetcar remained a major mode of transportation until the end of WWII. In the 1950’s, the system was dismantled and streetcar routes became bus routes. Though gaining in popularity from the 1930’s onward, the use of individual automobile really generalized after WWII, in parallel to a rapid development of the suburbs. The 1960’s saw a rapid development of major control access highways in the Island of Montreal (as well as the construction of a new Metro in the city center. Figure 39 illustrates how early

121

residential patches (i.e. whose average year span from origin to 1945) are mostly located in what could be termed the “inner city” with several exceptions of small patches scantily distributed in the Island. Residential patches of “intermediate” average year of creation (i.e. from 1946 to 1960) are mainly located in proximity to inner city area, or in suburbs created along the shores. More recently urbanized residential patches (i.e. average year from 1961 to present) are located in outer suburbs, in areas made easily accessible by control-access highways, as the city spread further and further. Comparing Figure 39 and Figure 34, the distribution of residential patches of different years and the distribution of different service levels are similar, which supports the above-mentioned result of negative correlation between year and access to fresh food. Table 25 shows the mean value of each variable in the three category of average year. All variables have highest mean values in the category of early average year and lowest mean values in the category of late.

122

Figure 39. Categories of average year of residential patches

Average year

Residential density variables Street connectivity variables

Coverage level Service level Average gross density Average net density Average number of nodes/ ha Average number of segments/ node Early [1896] - 1945 63.66 119.66 0.9297 1.92 68.40% 2.04 Intermediate 1946 - 1960 36.15 65.17 0.9187 1.78 46.91% 1.08 Late 1961 - present 28.34 43.92 0.7819 1.68 30.69% 0.46

Table 25. The mean values of all the variable in three categories of average year

Re-running the correlation analysis for patches of each of these groups of average years increases the level of confidence that the results actually measure the impact of morphological

123

features (pertaining to density and connectivity) on the presence and level of service offered by establishments selling fresh fruits and vegetables. Figures 40 - 42 illustrates the correlations between the variables in residential patches of three categories of average year.

The results illustrate that using the three time frames origin to 1945, 1946 to 1960, and 1961 to

now, allows to control for average year of urbanization, as in each category, the average year

does not correlate with other variables (to the exception of a weak correlation between year and

coverage level in the origin to 1945 category). Yet, both density variables as well as connectivity

variables are positively correlated with retail coverage level and retail service level in each time

category. Gross density in particular is strongly correlated with coverage level in category of intermediate average year, and is moderately correlated in categories of early and late average year. Net density is moderately correlated in each category. Number of street segments per node has a weak correlation with coverage and service levels in categories of intermediate and late average year, and has no correlation in category of early average year. Number of nodes per hectare has no correlation with coverage level and service level in category of intermediate average year, has weak correlation in category of late average year, and has weak correlation with coverage level but no correlation with service level in category of early average year. The service level and coverage level are moderately positively correlated (r = 0.424, 0.597) in categories of early and intermediate average year, and perfectly positively correlated (r = 0.932) in category of late average year. Gross and net densities display near perfect correlation (r = 0.968, 0.986, 0.985) in each category.

124

Figure 40. Correlation Coefficients (origin to 1945, N =73)

125

Figure 42. Correlation Coefficients (1961 to present, N =184)

The correlation analyses in patches corresponding to the three categories of income and three categories of average year have the similar results and findings with the global correlation analysis in all residential patches. When controlling for income and year, we can more confidently say that morphological variables paly a role in food access and service. In particular, residential density significantly affects both coverage and service levels.